Estimating Regional Wheat Yield By Integrating Remote Sensing Data With Simulation Model

Integrating of the real-time, regional remotely sensed data, the mathematical crop growth models can be applied at regional scale. This research could combine the advantages of the remote sensing and crop model, and overcome the lack of each technology. In this study, experiments with wheat including different varieties, nitrogen and planting density practices and a study area covered the whole Zhangjiagang City were conducted. The availability (timing and number) of remote sensing data were evaluated in the assimilating strategy. The strategy of integrating remotely sensed information with a wheat growth model by combining assimilation and updating strategies and the technology of regional wheat yield prediction based on simulation zones and coupling remotely sensed information with a wheat growth model were established. Basing on the prospective technology, the wheat growth monitoring and predicting system was designed and implemented for monitoring the wheat growth status and estimating regional wheat yield.A new strategy of integration of RS and crop growth model was developed for estimating wheat growth parameters and grain yield from local to regional areas by combing the assimilation and updating strategies which were used widely in the related researches. Leaf area index (LAI) and leaf nitrogen accumulation (LNA) were selected as the coupling factors which could be simulated by the crop model and retrieved by the remote sensing. The remotely sensed data ASD spectral, HJ-1A/B CCD data) from jointing to heading stage were used to reverse the three input parameters(sowing date, sowing rate and nitrogen rate) which could be used to optimize the simulation of the crop model by the assimilation strategies, then the remotly sensed datas of grain filling stage were used to revise the crop model by the updating strategies. The assimilation and updating strategies uses the optimization algorithm Particle Swarm Optimization (PSO) and deterministic algorithm Ensemble Square Root Filter (EnSRF) respectively. The integrated technique was test based on independent databases at field and regional scale, the results showed that the new approach could described the temporal and spatial distributions of wheat growth status and grain yields in the study area, with less than20%of the RE values for both growth parameter and regional total grain yield, Which indicated that the new strategies of integrating the crop model and remote sensing is promising for yield assessment scaled up from local to regional areas.When scaled up from field to regional areas, the technique of integrating the crop model and remote sensing faces a serious problem of the computational efficiency. Simulation zoning is using for regional agro-ecosystem modeling. Wheat canopy RVI (ratio vegetation index) data calculated from HJ-1A/B CCD images of three different winter wheat growth stages and soil nutrient indices, including total nitrogen content, organic matter content, available potassium content, were selected as data sources to study the zone division technique of wheat growth status in country level. With spatial variability, Fuzzy c-means a clustering algorithm was used to delineate management zones. The assimilation-updating strategy was applied to estimate the yield on the several zones respectively. The results showed that the calculation amount of integrating technology by simulation zones is more less than that of by pixels, and the values of variation coefficients of each zone divided by combined of RVI at heading were lower than the values of zone divided by individual RVI at jointing or grain filling, and RVIs at three stages performed better than RVI at single-stage. The grain yield of the each zone divided by RVIs at jointing, heading and grain filling stages were better homogeneous than that of the country level, and the yield estimated based on simulation zones was a little Higher than the yield based on pixels, more closer to the local statistics yield. The results would contribute to improve the computational efficiency of regional yield assessment technique of integrating the crop model and remote sensing.The RS-WheatGrow integration model based on integrating WheatGrow model and remotely sensed data was developed using object oriented programming technology. Microsoft.NET Framework3.5and Visual C#were selected as the development environment and the programming language to definite the system structure and interface. The system combines the crop growth model component; remote sensing processing modules developed by IDL; geographic information modules developed by ESRI ArcGIS Engine and the parallel PSO optimization algorithm module developed by GPU programming. With this system, varied functions could realized, such as image processing and spectral information extraction, growth monitoring based on RS, growth and yield simulating and predicting, thematic mapping.